From: Tom Lane <tgl@sss.pgh.pa.us>
To: "James Pang (chaolpan)" <chaolpan@cisco.com>
cc: PostgreSQL mailing lists <pgsql-bugs@lists.postgresql.org>
Subject: Re: huge memory of Postgresql backend process
In-reply-to: 
 <PH0PR11MB51911F681F37F55A92DA4C01D6409@PH0PR11MB5191.namprd11.prod.outlook.com>
References: 
 <PH0PR11MB51911F681F37F55A92DA4C01D6409@PH0PR11MB5191.namprd11.prod.outlook.com>
Comments: In-reply-to "James Pang (chaolpan)" <chaolpan@cisco.com>
	message dated "Thu, 08 Sep 2022 06:10:28 -0000"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <3950546.1662644877.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Thu, 08 Sep 2022 09:47:57 -0400
Message-ID: <3950547.1662644877@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/3950547.1662644877%40sss.pgh.pa.us>
Precedence: bulk

"James Pang (chaolpan)" <chaolpan@cisco.com> writes:
>   We just switch from Oracle to PGV13, all of these tables got partition=
 (256 HASH partition) ,  when UPDATE/DELETE ... WHERE . The memory increas=
e verify quickly until >130M. not too much data in these tables.  When we =
switch to PGV14.5 , test again, still see similar memory consumption issue=
.

I see no particular reason to think this is a bug.  If you have a
lot of partitions, there is going to be a lot of metadata for them.
Moreover, HASH partitioning is effectively incapable of being
pruned, so that every query is going to touch every partition.
(IMO, hash partitioning is basically never a good idea.)

			regards, tom lane